Research map / Video / Video segmentation and tracking
Mitigating Artifacts in Real-World Video Super-resolution Models
Liangbin Xie, Xintao Wang, Shuwei Shi, Jinjin Gu, Chao Dong, Ying Shan
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Replace RealBasicVSR with FastRealVSR | RealBasicVSR · VideoLQ | Runtime (ms): 63 -> 30 | Improved |
| Replace RealBasicVSR with FastRealVSR | RealBasicVSR · VideoLQ | NIQE: 3.7662 -> 3.7658 | Improved |
| Replace RealBasicVSR with FastRealVSR | RealBasicVSR · VideoLQ | BRISQUE: 29.030 -> 29.374 | Hurt |
| Replace RealBasicVSR with FastRealVSR | RealBasicVSR · VideoLQ | MANIQA (all frames): 0.4385 -> 0.4637 | Improved |
| Add both blurry and sharp hidden-state variants instead of using only blurry variants (ablation) | FastRealVSR · VideoLQ | NIQE: 3.8544 -> 3.7658 | Improved |
| Add both blurry and sharp hidden-state variants instead of using only sharp variants (ablation) | FastRealVSR · VideoLQ | NIQE: 3.8210 -> 3.7658 | Improved |
| Remove sharp hidden-state variants, retaining only blurry variants (ablation) | FastRealVSR · VideoLQ | NIQE: 3.7658 -> 3.8544 | Hurt |
| Remove blurry hidden-state variants, retaining only sharp variants (ablation) | FastRealVSR · VideoLQ | NIQE: 3.7658 -> 3.8210 | Hurt |
| Remove selective cross attention and use convolution to aggregate hidden states (ablation) | FastRealVSR · VideoLQ | NIQE: 3.7658 -> 3.8446 | Hurt |
| Use only a single blurry filter for the hidden-state pool (ablation) | FastRealVSR · VideoLQ | visual artifact suppression | Mixed |
| Use only a single sharp filter for the hidden-state pool (ablation) | FastRealVSR · VideoLQ | visual artifact suppression | Hurt |
Closest papers by meaning
- RealViformer: Investigating Attention for Real-World Video Super-Resolution (ECCV 2024 · similarity 0.90)
- Investigating Tradeoffs in Real-World Video Super-Resolution (CVPR 2022 · similarity 0.90)
- SkipSR: Faster Super-Resolution with Token Skipping (NeurIPS 2026 · similarity 0.89)
- Memory-Augmented Non-Local Attention for Video Super-Resolution (CVPR 2022 · similarity 0.88)
- One-Step Diffusion for Detail-Rich and Temporally Consistent Video Super-Resolution (NeurIPS 2025 · similarity 0.87)
- NegVSR: Augmenting Negatives for Generalized Noise Modeling in Real-world Video Super-Resolution (AAAI 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
